Real-Time Protocol/Real-Time Control Protocol
The Real-Time Protocol (RTP) and Real-Time Control Protocol (RTCP) were created to transport real-time traffic, such as video and audio, over the Internet and/or Internet
Protocol–based networks. RTP is a session-layer protocol that usually runs on top of UDP/IP and is defined in RFCs 1889 and 1890. UDP is the favored transport-layer protocol because its connectionless nature enables its use in broadcast and multicast environments. RTP addresses some of the elements required for real-time traffic which are missing from UDP, including sequence numbering and time stamping. Sequence numbering enables the receiver to verify that datagrams have been received in order and that intermediate datagrams were not lost. Time stamping helps ensure proper playback rates regardless of datagram arrival times. Additionally, a payload-type field helps the receiving station identify which application or process to hand off the data to. Given this functionality Cisco and most of the VoIP community have chosen to use RTP for the transport of packetized voice in IP networks.
RTCP carries out the control functions for RTP streams. RTCP performs four discrete functions:
1. Communication of information and statistics about the RTP stream to the application
3. Limiting of control traffic
4. Secondary transport for small amounts of information
This protocol is important in voice environments for its ability to identify RTP sources and for the statistical information it provides applications.
Resource Reservation Protocol
Resource Reservation Protocol, or RSVP, is a protocol developed to allow applications to dynamically request specific levels of service from the network. RSVP's application-based approach enables it to be customized based on specific application requirements. RSVP-enabled applications communicate their bandwidth and network latency requirements to receiving applications. The receivers then issue RSVP requests to the network elements along the return path to the source station for these resources. Once a network element agrees to support a request, it is expected to honor that request for the duration of the reservation. In voice
Page 23 over IP environments, a voice-enabled router can issue RSVP reservations for each individual voice call and rely on the intermediate network elements to dynamically allocate the appropriate resources for the call.
In addition to standard RSVP requests over packet networks, Cisco offers a means to map RSVP reservations to ATM SVCs called RSVP-ATM QoS interworking. This feature is supported by the 7500 series and enables the router to dynamically set up an ATM SVC with the appropriate QoS parameters to support the RSVP request. In hybrid packet and cell environments,
RSVP-ATM QoS interworking provides enhanced QoS integration and leverages ATM's intrinsic QoS capabilities to support packet-based requests.
Diffserv/Committed Access Rate
Cisco's committed access rate (CAR) and the goals of the IETF's differentiated services (diffserv) working group are similar. The diffserv group is developing a standards-based approach to classifying IP traffic, setting bits within the IP header to identify a specific QoS, providing preferential service based on QoS levels, and supporting the QoS levels throughout the network. Cisco's committed access rate is an early implementation that provides rate control for IP traffic, interpretation and setting of IP precedence bits, and differentiated service throughout the network based upon IP precedence bits. This feature is significant in voice environments as it helps provide network-wide classes of service for IP traffic.
H.323
H.323 is an International Telecommunications Union (ITU) standard which was created in 1996 and updated in 1998. It provides a foundation for audio, video, and data communications across a packet-based network infrastructure. H.323 provides standards for voice-encoding, simple
and links to external networks. The voice over IP community has adopted H.323 standards in an effort to foster the interoperability of equipment from multiple vendors. H.323 provides a solid foundation for building multimedia networks and has become critical for developing large-scale VoIP networks.
G.729/G.729a
G.729 and G.729a are standard for voice-encoding algorithms. These algorithms analyze standard PCM voice segments and use a complex algorithm to encode them. Once encoded, a lookup pointer for the audio
Page 24 pattern's location in a shared code dictionary is forwarded to the receiving node. Since the actual audio signal is not sent over the network (just the code), this algorithm results in significant bandwidth savings. G.729 and G.729a are particularly significant in the voice over IP arena because their low-bandwidth requirements (8 kbps) help to offset the overhead introduced by encapsulating voice traffic in data packets.
Session Initial Protocol
The Session Initiation Protocol, or SIP, is the IETF initiative for providing audio and/or video transport over the Internet. It is an application-layer control (signaling) protocol for creating, modifying, and terminating sessions with one or more participants. These sessions include Internet multimedia conferences, Internet telephone calls, and multimedia distribution. Members in a session can communicate via multicast tree or a mesh of unicast relations. SIP is defined by RFC 2543 and is being considered by a number of vendors and service providers.
Packet over SONET
Packet over SONET provides higher speed, more efficient transmission of IP traffic across service provider backbones. Given that most of the high-speed links deployed are point-to-point links between routers, the signaling overhead and cell tax associated with ATM is an unnecessary inefficiency. Packet over SONET maps PPP or HDLC frames directly into SONET frames
avoiding the segmentation and reassembly processes associated with ATM. This improves efficiency significantly by increasing the effective throughput over an OC-3c link from approximately 135 Mbps to over 155 Mbps. Both the performance gains and ease of
implementation have made Packet over SONET a popular choice for large-scale IP networks. Packet over SONET is significant in voice environments in that it provides a more efficient way to transfer IP packets across network backbones.
FRF.11/FRF.12
The Frame-Relay Forum's FRF.11 specifies an implementation of voice over frame relay, while FRF.12 specifies a method for segmentation and reassembly of frame-relay frames at the
data-link (frame) level. FRF.12 is somewhat analogous to ATM's cell structure in that it provides a maximum frame size for transmission through the network to help control delay and variability
in frame arrival rates. It differs from ATM in that frame size is not fixed and varies with the link speeds involved. FRF.11
Page 25 support is important because it provides interoperability not only between Cisco's MC3810 and Cisco 2600/3600 products, but between these products and products from other vendors which are FRF.11 compliant. FRF.12 is significant in both voice over frame relay and voice over IP environments as it specifies a means for limiting the delay and variability introduced by variable-length frames.
Multilink PPP
A subset of the Multilink PPP specification includes link fragmenting and interleaving, or LFI. LFI allows a single frame to split up and simultaneously transmit over multiple links in the multilink bundle yielding improved utilization across all links and reducing transmission delays. The far-end router receives the frame segments, reassembles them, and routes the complete frame as appropriate. This function is also used to help reduce latency and prioritize voice traffic in voice over IP networks. In voice networks, LFI helps bound the transmission delay by splitting large frames into smaller segments with a well-defined transmission delay. To ensure that voice traffic is serviced appropriately, a separate queue is created for voice traffic and given priority access to the network link. Delay is bounded by injecting voice frames in-between the segments of larger frames. Multilink PPP's LFI feature can be enabled over single PPP links, as well as multilink bundles.